Pritom Saha | Searcharxiv

arXiv2026

The First Human-Based In Vitro Flow Loop and Quantification for Fetal Aortic Hemodynamics

Coarctation of the aorta (CoA) is a common congenital defect that remains difficult to diagnose prenatally due to subtle and evolving anatomical features. In the fetus, the ductus arteriosus creates a dual-inflow configuration that generates complex three-dimensional flow patterns not captured by standard imaging. Improved characterization of fetal hemodynamics may enhance diagnostic accuracy beyond anatomy-based assessment. This study presents the first human-based in vitro flow loop of the fetal aorta, constructed from anatomies reconstructed using medical imaging data. Models representing normal and coarctation conditions were fabricated and integrated into a physiological flow loop. Velocity fields were measured using planar and stereoscopic particle image velocimetry (PIV) to resolve near-wall and three-dimensional flow structures, enabling quantitative assessment of velocity gradients and wall shear stress (WSS) under normal and coarctation configurations. The in vitro flow loop closely reproduced target fetal flow segmentation, with segmental flow-rate errors generally below 6%. High-resolution planar and stereoscopic PIV revealed dual jets from the ascending aorta and the ductus arteriosus and predominantly planar flow in the normal aorta, but strong jet acceleration, separation, and reattachment in the coarcted geometry. Coarctation produced markedly elevated and spatially heterogeneous WSS, and 2-component PIV underestimated WSS by up to ~29% compared with 3-component measurements, especially in high-shear regions. These findings show that accurate three-component velocity measurements are critical for reliable WSS estimation and suggest that detailed hemodynamic metrics, such as WSS, may serve as potential biomarkers to enhance fetal CoA diagnosis beyond anatomy alone.

physics.med-ph↗

arXiv2022

Rethinking Task-Incremental Learning Baselines

It is common to have continuous streams of new data that need to be introduced in the system in real-world applications. The model needs to learn newly added capabilities (future tasks) while retaining the old knowledge (past tasks). Incremental learning has recently become increasingly appealing for this problem. Task-incremental learning is a kind of incremental learning where task identity of newly included task (a set of classes) remains known during inference. A common goal of task-incremental methods is to design a network that can operate on minimal size, maintaining decent performance. To manage the stability-plasticity dilemma, different methods utilize replay memory of past tasks, specialized hardware, regularization monitoring etc. However, these methods are still less memory efficient in terms of architecture growth or input data costs. In this study, we present a simple yet effective adjustment network (SAN) for task incremental learning that achieves near state-of-the-art performance while using minimal architectural size without using memory instances compared to previous state-of-the-art approaches. We investigate this approach on both 3D point cloud object (ModelNet40) and 2D image (CIFAR10, CIFAR100, MiniImageNet, MNIST, PermutedMNIST, notMNIST, SVHN, and FashionMNIST) recognition tasks and establish a strong baseline result for a fair comparison with existing methods. On both 2D and 3D domains, we also observe that SAN is primarily unaffected by different task orders in a task-incremental setting.

cs.AI↗